RFE: some way of smoothing molecular surface cusps

[Elaine Meng]

Sometimes I get cusps on the molecular surface which I assume may be mathematically correct, but don’t look so nice… and I was wondering if you have any ideas of how to avoid them.

In this case (5ego) I don’t have the problem with the DNA surfaces calculated separately for each chain (attached image smooth.png), but when I try to make a single surface enclosing both chains, the sharp points appear (attached image pointy.png). Changing probe radius and/or grid spacing does not help. I wish for some way to smooth or remove them, even though these cusps may be mathetically correct for the input geometry.

However, I did the experiment of generating a grid surface in Chimera1 (session attached pointygridsurf.py) and then trying "sop smooth" but that method was not very successful. Seems like the rest of the surface gets smoothed more than the points.

[Tom Goddard]

Unfortunately the "sop smooth" command in Chimera that moves vertices towards their neighbor vertices does not work well if the triangle mesh is not uniformly spaced. Often a cusp has many tightly spaced vertices and the smoothing has less of an effect on those cusps then on the surrounding already smooth surface. So if you increase the smoothing parameters enough to smooth the cusp, then the rest of the surface has been dramatically changed even though it was already smooth.

I've attached a picture of the mesh of a cusp of the example pointy surface showing the closely spaced vertices.

It might be by using "sop finerMesh" to make the whole surface have a mesh of similar fineness to the cusp, then using sop smooth could get a decent result. But I'm skeptical.

Problems of altering surfaces with bad triangulation (irregular size triangles) are very difficult and Chimera has no capability to deal with this -- it is a whole field of computer science.

[Tom Goddard]

Mesh around a cusp shows tightly spaced vertices that make it resist smoothing.

[Elaine Meng]

Current best solution is to decrease probe radius even more than I tried before. Going down to probe radius 1.1 gets rid of the cusps in the 5ego "enclose nucleic" surface. I also tried Gaussian surface but that changed the shape a lot more than simply decreasing SES probe radius.